PASSCoDe: Parallel ASynchronous Stochastic dual Co-ordinate Descent

TL;DR

PASSCoDe introduces asynchronous parallel stochastic dual coordinate descent algorithms that significantly accelerate solving large-scale regularized empirical risk minimization problems on multi-core systems.

Contribution

It proposes a novel asynchronous parallel SDCD algorithm with convergence analysis and demonstrates superior speedup over existing methods.

Findings

Achieves linear convergence with atomic operations.

First backward error analysis for ASDCD without locking.

Significantly faster than previous parallel solvers.

Abstract

Stochastic Dual Coordinate Descent (SDCD) has become one of the most efficient ways to solve the family of $\ell_2$-regularized empirical risk minimization problems, including linear SVM, logistic regression, and many others. The vanilla implementation of DCD is quite slow; however, by maintaining primal variables while updating dual variables, the time complexity of SDCD can be significantly reduced. Such a strategy forms the core algorithm in the widely-used LIBLINEAR package. In this paper, we parallelize the SDCD algorithms in LIBLINEAR. In recent research, several synchronized parallel SDCD algorithms have been proposed, however, they fail to achieve good speedup in the shared memory multi-core setting. In this paper, we propose a family of asynchronous stochastic dual coordinate descent algorithms (ASDCD). Each thread repeatedly selects a random dual variable and conducts coordinate updates using the primal variables that are stored in the shared memory. We analyze the convergence properties when different locking/atomic mechanisms are applied. For implementation with atomic operations, we show linear convergence under mild conditions. For implementation without any atomic operations or locking, we present the first {\it backward error analysis} for ASDCD under the multi-core environment, showing that the converged solution is the exact solution for a primal problem with perturbed regularizer. Experimental results show that our methods are much faster than previous parallel coordinate descent solvers.